Level: Beginner
In non-clinical trials in the pharmaceutical industry, there are many cases where the efficacy of a drug is proven by hypothesis testing in a test with multiple drug doses, and multiple comparisons are essential to strictly control type I error. In particular, the Dunnett multiple comparison method is widely used to compare a control group with multiple drug dose groups, but there is a problem that the statistical power decreases in tests with a large number of groups in order to simultaneously compare the control group with each dose.
The Williams multiple comparison method is applied when the drug effect is dose-dependent (monotonic). Since sequential tests are performed in a closed procedure starting from the high-dose group, it is an attractive multiple comparison method with almost no loss in statistical power compared to the two-group t-test. Unfortunately, however, it is not currently implemented in JMP.
Therefore, we are developing an add-in that uses JSL to perform sequential testing in a closed procedure starting from the high-dose group based on the Williams method, compares the calculated statistics with the statistical table for Williams multiple comparisons in a textbook, and outputs whether there is a significant difference. We are also planning to include the non-parametric Shirley-Williams multiple comparison method. In the presentation, we will introduce the functions of the add-in, including a demo of the prototype.
Since last year, he has been a systems engineer in charge of pharmaceutical-related non-clinical business at Takumi Information Technology Co., Ltd. He is engaged in developing statistical analysis software for non-clinical departments, and lecturing and consulting at statistical seminars.
He has been engaged in in vitro and in vivo efficacy pharmacology testing in the pharmacology department of a pharmaceutical company for many years, and has experience in applying for manufacturing and marketing approval. In 2007, he transferred to the research management department of a group company and was in charge of non-clinical statistical analysis, where he was taught JMP by Dr. Toshiro Haga and Dr. Yukio Takahashi, and has continued to work there ever since. Recently, he has been mainly in charge of development using JMP's scripting language (JSL), and is planning to create a series of non-clinical add-ins that expand JMP functions.
